headers
Andrew Beekhof | 2 Sep 2008 21:29
Picon

A note about the upcoming 1.0 release

In approximately 2 weeks, I will be creating a Stable-1.0 branch from  
whatever is in 'dev' at the time.  After that I will begin release  
testing and, if experience is any guide, 1.0.0 should be ready by the  
first week of November.

This means that there is no better time to check out 0.7 (aka. dev)  
and explore how the new (and old) features work - because once the  
code is frozen^, that's how they're going to work for the next couple  
of years.

Please do take the time to test drive the new code.  There really are  
some great enhancements that make creating and maintaining the cluster  
a great deal easier than before.  I'm sure once you try it, you'll not  
want to go back to 0.6 ;-)

Also, be sure to check out the updated, and much expanded,  
documentation.  You can download the PDF from:
    http://clusterlabs.org/mw/Image:Configuration_Explained_1.0.pdf

regards,
Andrew

^ Its practically frozen already.  There are at most 2 new features to  
go in (but they'll probably not be ready in time) and only your  
feedback and bug reports will be cause for any last minute changes.
Permalink | Reply | 10 comments |
headers
Raoul Bhatia [IPAX] | 4 Sep 2008 11:45
Picon
Favicon
Gravatar

Re: A note about the upcoming 1.0 release

Andrew Beekhof wrote:
> In approximately 2 weeks, I will be creating a Stable-1.0 branch from
> whatever is in 'dev' at the time.  After that I will begin release
> testing and, if experience is any guide, 1.0.0 should be ready by the
> first week of November.
> 
> 
> This means that there is no better time to check out 0.7 (aka. dev) and
> explore how the new (and old) features work - because once the code is
> frozen^, that's how they're going to work for the next couple of years.
> 
> 
> Please do take the time to test drive the new code.  There really are
> some great enhancements that make creating and maintaining the cluster a
> great deal easier than before.  I'm sure once you try it, you'll not
> want to go back to 0.6 ;-)

is there any possibility to get pre-packaged 0.7.x packages from the
opensuse buildservice? i would need it for debian etch.

cheers,
raoul
--

-- 
____________________________________________________________________
DI (FH) Raoul Bhatia M.Sc.          email.          r.bhatia@...
Technischer Leiter

IPAX - Aloy Bhatia Hava OEG         web.          http://www.ipax.at
Barawitzkagasse 10/2/2/11           email.            office@...
1190 Wien                           tel.               +43 1 3670030
(Continue reading)

Permalink | Reply | 7 comments |
headers
Junko IKEDA | 4 Sep 2008 12:00
Picon

score configuration to node

Hi,

I usually set <constraints> tag like this,

<constraints>
  <rsc_location id="rsc_location" rsc="dummy">
    <rule id="rule1" score="200">
       <expression id="exp1" attribute="#uname" operation="eq"
value="node-a"/>
    </rule>
    <rule id="rule2" score="100">
       <expression id="exp2" attribute="#uname" operation="eq"
value="node-b"/>
    </rule>
  </rsc_location>
</constraints>

in this case, resource will be started preferentially on node-a.
If I don't set the score for node, default value would be zero, right?
I want set its default as -INFINITY, it means that non-listed node can not
have any resources.
Is this possible?
If I can set this value as "cluster_property_set", it would be very
convenient.

Best Regards,
Junko Ikeda

NTT DATA INTELLILINK CORPORATION
(Continue reading)

Permalink | Reply | 4 comments |
headers
Lars Marowsky-Bree | 4 Sep 2008 12:04
Picon

Re: score configuration to node

On 2008-09-04T19:00:05, Junko IKEDA <ikedaj@...> wrote:

> in this case, resource will be started preferentially on node-a.
> If I don't set the score for node, default value would be zero, right?
> I want set its default as -INFINITY, it means that non-listed node can not
> have any resources.
> Is this possible?

The default value is determined by the "symmetric-cluster" property.

Regards,
    Lars

--

-- 
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
Permalink | Reply | 3 comments |
headers
Junko IKEDA | 4 Sep 2008 12:17
Picon

RE: score configuration to node

> > in this case, resource will be started preferentially on node-a.
> > If I don't set the score for node, default value would be zero, right?
> > I want set its default as -INFINITY, it means that non-listed node can
not
> > have any resources.
> > Is this possible?
> 
> The default value is determined by the "symmetric-cluster" property.

Thanks!
With my quick check, it worked well.
When I have a lot of nodes, I will be able to cut off the size of cib.xml
with this configuration.

Thanks,
Junko
Permalink | Reply | 2 comments |
headers
Andrew Beekhof | 4 Sep 2008 14:09
Picon

Re: A note about the upcoming 1.0 release


On Sep 4, 2008, at 11:45 AM, Raoul Bhatia [IPAX] wrote:

> Andrew Beekhof wrote:
>> In approximately 2 weeks, I will be creating a Stable-1.0 branch from
>> whatever is in 'dev' at the time.  After that I will begin release
>> testing and, if experience is any guide, 1.0.0 should be ready by the
>> first week of November.
>>
>>
>> This means that there is no better time to check out 0.7 (aka. dev)  
>> and
>> explore how the new (and old) features work - because once the code  
>> is
>> frozen^, that's how they're going to work for the next couple of  
>> years.
>>
>>
>> Please do take the time to test drive the new code.  There really are
>> some great enhancements that make creating and maintaining the  
>> cluster a
>> great deal easier than before.  I'm sure once you try it, you'll not
>> want to go back to 0.6 ;-)
>
> is there any possibility to get pre-packaged 0.7.x packages from the
> opensuse buildservice? i would need it for debian etch.

Of course, they've been around for while :-)

http://download.opensuse.org/repositories/server:/ha-clustering:/UNSTABLE/
(Continue reading)

Permalink | Reply | 6 comments |
headers
Junko IKEDA | 5 Sep 2008 07:42
Picon

RE: score configuration to node

> > The default value is determined by the "symmetric-cluster" property.
> 
> Thanks!
> With my quick check, it worked well.
> When I have a lot of nodes, I will be able to cut off the size of cib.xml
> with this configuration.

We could cut off the size of cib.xml using symmetric-cluster="false",
but there is the other problem...
Probes actions run on the all node, so the size of "cibadmin -Q" would get
bigger and bigger if the number of node is increased.
Is it possible to restrain the probe action, if the score for one node set
as -INFINITY?
Probe for node would be needed, but probe for resources which will never run
on that node might be unnecessary.
We have 9 nodes, (8 actives and 1 standby) and each node has 16 resources.
Heartbeat version is 2.1.3.

Thanks,
Junko
Permalink | Reply | 1 comment |
headers
Dejan Muhamedagic | 5 Sep 2008 16:59

the crm command line interface

Hi,

Can't read XML? Then this may be something for you...

For some time the command line interface for the CRM/Pacemaker
has been in the works. About one month ago I posted a draft
specification and user guide to the development list. There was
no response, so I assumed that everybody was happy with it ;-)

As I got better grip on how the program could be used, the
specification changed accordingly during the development.

The first release is already in the Pacemaker's development
repository. You can get it here:

http://hg.clusterlabs.org/pacemaker/dev/raw-file/tip/tools/crm.in

To run it, first replace  <at> HA_VARLIBDIR <at>  with /var/lib or wherever
heartbeat/crm lives. Or compile and install the latest pacemaker
code. Andrew will probably soon make new rpms.

The source of the current documentation:

http://hg.clusterlabs.org/pacemaker/dev/raw-file/tip/doc/crm_cli.txt

It is readable, but to produce an esthetically more pleasing
document, just process it with asciidoc to get PDF or HTML.

CLI consists of two main parts: the management (resource
start/stop, node standby, and similar) and the configuration
(Continue reading)

Permalink | Reply | 0 comments |
headers
Serge Dubrouski | 5 Sep 2008 17:42
Picon

Re: [Linux-HA] the crm command line interface

Hello -

It's a really great tool. Great job! Here are my expirience
(pacemaker-0.6.5-2.2 + Heartbeat 2.1.3-23.1:

Nodes.

1. Status and Show Node show the exactly the same information. Even
when you provide node name in show command it shows all nodes. Is it
supposed so?

cib.

Couldn't test it because crm_shadow wasn't found. Looks like it's a new tool.

Resource.

1. Status without parameters shows a list of resources but doesn't
show the real status of them onlinr/stopped.

2. It's not clear how to migrate a resource to a specific node it it's
possible at all.

3. It would be good if param resource show was able to show all
parameters for a given resource.

4. Again would be good failcount resource was able to show failcounts
for all nodes

Configure.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Dejan Muhamedagic | 5 Sep 2008 18:38

Re: Re: [Linux-HA] the crm command line interface

Hi Serge,

On Fri, Sep 05, 2008 at 09:42:41AM -0600, Serge Dubrouski wrote:
> Hello -
> 
> It's a really great tool. Great job!

Thanks.

> Here are my expirience
> (pacemaker-0.6.5-2.2 + Heartbeat 2.1.3-23.1:
> 
> Nodes.
> 
> 1. Status and Show Node show the exactly the same information. Even
> when you provide node name in show command it shows all nodes. Is it
> supposed so?

No. Actually, I'm looking for input here. Don't exactly know
what's interesting for users. There are two kinds of information:
provided by the CIB (node type and node attributes) and by
crmadmin (if the node's online and if it is DC).

> cib.
> 
> Couldn't test it because crm_shadow wasn't found. Looks like it's a new tool.

Yes. That's a very useful new thing provided by Andrew. It
basically allows you to test the configuration out of the
cluster.
(Continue reading)

Permalink | Reply | 0 comments |

Gmane